don't rely on std:: overloads of arg, conj, imag, and real for real args * lo-mappers.h (arg, conj, imag, real): Provide overloads for real arguments in octave::math namespace instead of importing them from std namespace. * __ichol__.cc, ls-mat4.cc, ls-mat5.cc, __eigs__.cc, ov-complex.cc, ov-complex.h, ov-cx-mat.cc, ov-cx-sparse.cc, ov-flt-complex.cc, ov-flt-complex.h, ov-flt-cx-mat.cc, CMatrix.cc, CNDArray.cc, CSparse.cc, MatrixType.cc, dMatrix.cc, dSparse.cc, fCMatrix.cc, fCNDArray.cc, fMatrix.cc, chol.cc, eigs-base.cc, lo-mappers.cc, lo-specfun.cc, sparse-qr.cc: Update for arg, conj, imag, and real functions in octave::math namespace. Prefer member function versions of real and imag for simple variables. Use std:: namespace versions directly for known complex values. Use octave::math namespace versions otherwise

--- a/liboctave/array/CNDArray.cc	Fri Aug 12 00:19:11 2016 +0100
+++ b/liboctave/array/CNDArray.cc	Thu Aug 11 16:26:56 2016 -0400
@@ -540,8 +540,8 @@
     {
       Complex val = elem (0);
 
-      double r_val = std::real (val);
-      double i_val = std::imag (val);
+      double r_val = val.real ();
+      double i_val = val.imag ();
 
       max_val = r_val;
       min_val = r_val;
@@ -559,8 +559,8 @@
     {
       Complex val = elem (i);
 
-      double r_val = std::real (val);
-      double i_val = std::imag (val);
+      double r_val = val.real ();
+      double i_val = val.imag ();
 
       if (r_val > max_val)
         max_val = r_val;